R Ch
R Ch
Sep 5, 2018 · 2 min read

Pacman

Since I feel I missed out on taking a cool AI class during college days, I figured I could compensate for it by taking one close to a decade later.

The interactive courses are the ones that tend to stick with me the most, so the Berkeley AI course: http://ai.berkeley.edu/ feels like a great start. Even though this course is kind of old (2014), I think it’s a good start to get a grasp on the fundamentals of AI.

The Berkeley AI course uses Pacman to explore several important topics within AI. There are several subcomponents to make it easy to tackle small concepts and there is an autograder that makes it easy to grade on my own.

My goals for this course are to widen the depth of my AI understanding. To this end, I aim to go deep down in every aspect, even if it means the course takes a lot longer to complete than I expected. I am tempted to put timelines but know they rarely work out, and are counter-productive to learning in-depth content.

Through the medium of this blog, I hope to track in the following:

  • Watching videos of the class and summarizing things I learnt in it
  • Critical reviews — As a beginner, I of course don’t consider myself worthy of reviewing people who are experts in the field. However, I can relate to how other beginners may find certain explanations difficult to understand coming from a similar background, and what would have been more helpful. Conversely, I will point out explanations that helped me something new that I didn’t know before.
  • Completing assignments — Since I won’t be having the help of TAs or classmates, it is inevitable that I will get stuck. I will permit myself to access online resources to get around these. I will blog any parts that I think were in my knowledge blind spots prior to the assignment. I am sure there would still be parts that I don’t understand, I will make a note of these as open questions and update them if I learn of the answers.
  • Broader applications for what I learnt — getting into the nitty-gritties of things it is easy to forget the broader applications of what is being learnt, I hope to use this blog to remind me to think of those
  • Exams — While I’m not a big fan of rote learning, I do think exams surface the gaps of knowledge and expedite learning better than several other mediums. So for every small chunk I tackle, I will attempt some of the corresponding exam questions to get a more holistic understanding of the topics.

Each blog would cover one tiny aspect of the Pacman built and some intermediate posts would tie these pieces together.

I hope to use this blog to be a learning medium that is technical yet easy to understand for someone new to the field and in the process forces me to be accountable for completing the course.